As a technology of identifying whether or not a desired object exists in a certain space, the marker detection method of previously adding a marker with a predetermined shape to a targeted object, and determining whether or not the targeted object exists, dependent upon existence or not of the marker in a background video image of the above space, is generally known.
For example, the technology of extracting feature points from the background video image not containing the marker, and generating the marker to be derived based on the above extracted feature points, which is comprised of image features not appearing in the background video image, is disclosed in Patent literature 1.
Specifically, when a pattern identical to the marker generated in such a method is previously added to the targeted object in a preparatory stage, and this pattern is detected from a predetermined space video image in a detection stage, it may be determined that the targeted object exists in the above space.
As it is, in a detection stage, the pattern of the marker added to the object could be varied in some cases, accompanied with a change in the environments such as a direction of a camera for imaging the above space and the like.
This causes a problem that the pattern of the marker added to the object does not match with the pattern of the marker that should be detected, and the marker cannot be detected appropriately.
Thereupon, the technology of previously storing a first geometric invariant corresponding to an object identification information code (that corresponds to the aforementioned marker) peculiar to the object for each object in the preparatory stage, acquiring the images of the object added with the object identification information code from a plurality of viewpoints in the detection stage, extracting feature point positions from a plurality of independent images produced out of these images to obtain a second geometric invariant, and detecting the first geometric invariant being the closest to this second geometric invariant, thereby to detect the corresponding object is disclosed in Patent literature 2.
That is, in accordance with such a technology, making a comparison of the object identification information based upon the invariant without depending upon a plurality of the viewpoints makes it possible to identify identity between the object identification information acquired in a certain direction and the object identification information acquired in another direction.
For this, the above-mentioned problems can be dissolved because the targeted object can be accurately detected even though the video images are acquired obliquely.